Eurographics Digital Library
This is the DSpace 7 platform of the Eurographics Digital Library.
- The contents of the Eurographics Digital Library Archive are freely accessible. Only access to the full-text documents of the journal Computer Graphics Forum (joint property of Wiley and Eurographics) is restricted to Eurographics members, people from institutions who have an Institutional Membership at Eurographics, or users of the TIB Hannover. On the item pages you will find so-called purchase links to the TIB Hannover.
- As a Eurographics member, you can log in with your email address and password from https://services.eg.org. If you are part of an institutional member and you are on a computer with a Eurographics registered IP domain, you can proceed immediately.
- From 2022, all new releases published by Eurographics will be licensed under Creative Commons. Publishing with Eurographics is Plan-S compliant. Please visit Eurographics Licensing and Open Access Policy for more details.
Recent Submissions
The Misclassification Likelihood Matrix: Some Classes Are More Likely To Be Misclassified Than Others
(The Eurographics Association, 2024) Sikar, Daniel; Garcez, Artur d'Avila; Bloomfield, Robin; Weyde, Tillman; Peeroo, Kaleem; Singh, Naman; Hutchinson, Maeve; Laksono, Dany; Reljan-Delaney, Mirela; Hunter, David; Slingsby, Aidan
This study introduces the Misclassification Likelihood Matrix (MLM) as a novel tool for quantifying the reliability of neural network predictions under distribution shifts. The MLM is obtained by leveraging softmax outputs and clustering techniques to measure the distances between the predictions of a trained neural network and class centroids. By analyzing these distances, the MLM provides a comprehensive view of the model's misclassification tendencies, enabling decision-makers to identify the most common and critical sources of errors. The MLM allows for the prioritization of model improvements and the establishment of decision thresholds based on acceptable risk levels. The approach is evaluated on the MNIST dataset using a Convolutional Neural Network (CNN) and a perturbed version of the dataset to simulate distribution shifts. The results demonstrate the effectiveness of the MLM in assessing the reliability of predictions and highlight its potential in enhancing the interpretability and risk mitigation capabilities of neural networks. The implications of this work extend beyond image classification, with ongoing applications in autonomous systems, such as self-driving cars, to improve the safety and reliability of decision-making in complex, real-world environments.
Investigation of the Organisation of the Extracellular Matrix Using Fibre Tracing
(The Eurographics Association, 2024) Arafat, Youssef; Cuesta-Apausa, Cristina; Castellano, Esther; Reyes-Aldasoro, Constantino Carlos; Hunter, David; Slingsby, Aidan
Throughout history, the observation of medical and biological samples has been of high importance and has led to many discoveries. When this process relies on human observation, it can be time-consuming, especially with the advent of technological advancements that generate more and more images at faster rates. Additionally, some features of the samples can be undetectable by the naked eye, but with the aid of visual computing techniques, these hidden details can be revealed. The morphological characteristics of the extracellular matrix play a vital role in cancer and other health conditions. Visual observations of the ECM can provide valuable insights; however, the task may be tedious and sometimes it is hard to quantify the differences between samples. In this work, a tracing algorithm is proposed. Furthermore, morphological characteristics of the extracellular matrix can be extracted with the algorithm to quantify and compare different biological populations. Experiments revealed that the removal of interactions in fibroblasts affected their ability to form a healthy extracellular matrix as compared with a wild type population. Here, an investigation of the morphological differences between the ECM of two populations was conducted. Five images of mutant and five images of wild type cells growing in culture were compared. A deconvolutional convolutional neural network was used as a pre-processing filtering method to remove noise from the images. The images are then traced by the proposed algorithm, Trace Ridges, to extract morphological features and visually present the edges and gaps extracted. Trace Ridges combines methods of Edge detection, watershed, and morphological characteristics to delineate fibre-like structures. Two morphological characteristics provided statistical differences between the populations: number of fibres (p−value = 0.00091) and relative area of gaps between the fibres (p−value = 0.014). The number of fibres detected in wild type was higher than mutant while the relative gaps area size of mutant was higher than that of WT. Trace Ridges was able to successfully delineate the ECM fibres of mutant and wild type cells and extract morphological features to show the difference between the populations.
LLM-Assisted Visual Analytics: Opportunities and Challenges
(The Eurographics Association, 2024) Hutchinson, Maeve; Jianu, Radu; Slingsby, Aidan; Madhyastha, Pranava; Hunter, David; Slingsby, Aidan
We explore the integration of large language models (LLMs) into visual analytics (VA) systems to transform their capabilities through intuitive natural language interactions. We survey current research directions in this emerging field, examining how LLMs are integrated into data management, language interaction, visualisation generation, and language generation processes. We highlight the new possibilities that LLMs bring to VA, especially how they can change VA processes beyond the usual use cases. We especially highlight building new visualisation-language models, allowing access of a breadth of domain knowledge, multimodal interaction, and opportunities with guidance. Finally, we carefully consider the prominent challenges of using current LLMs in VA tasks. Our discussions in this paper aim to guide future researchers working on LLM-assisted VA systems and help them navigate common obstacles when developing these systems.
Visual Interpretation of Tagging: Advancing Understanding in Task-Oriented Dialogue Systems
(The Eurographics Association, 2024) Zhou, Yazhuo; Xing, Yiwen; Abdul-Rahman, Alfie; Borgo, Rita; Hunter, David; Slingsby, Aidan
In task-oriented dialogue systems, tagging tasks leverage Large Language Models (LLMs) to understand dialogue semantics. The specifics of how these models capture and utilize dialogue semantics for decision-making remain unclear. Unlike binary or multi-classification, tagging involves complex multi-to-multi relationships between features and predictions, complicating attribution analyses. To address these challenges, we introduce a novel interactive visualization system that enhances understanding of dialogue semantics through attribution analysis. Our system offers a multi-level and layer-wise visualization framework, revealing the evolution of attributions across layers and allowing users to interactively probe attributions. With a dual-view for streamlined comparisons, users can effectively compare different LLMs. We demonstrate our system's effectiveness with a common task-oriented dialogue task: slot filling. This tool aids NLP experts in understanding attributions, diagnosing models, and advancing dialogue understanding development by identifying potential sources of model hallucinations.
Min-Max Modifiable Nested Octrees M3NO: Indexing Point Clouds with Arbitrary Attributes in Real Time
(The Eurographics Association, 2024) Hermann, Paul; Krämer, Michel; Dorra, Tobias; Kuijper, Arjan; Hunter, David; Slingsby, Aidan
We present a data structure that allows 3D point clouds with arbitrary attributes to be indexed in real time. We focus on large datsets from mobile mapping systems such as airborne and terrestrial laser scanners. Compared to traditional indexing approaches running offline, our data structure can be created incrementally while the points are being recorded. This allows the data to be used (i.e. analyzed or visualized) already during acquisition or immediately after it has finished. The data structure enables queries based on spatial extent and value ranges of arbitrary attributes. This is in contrast to existing works, which focus on either spatial or attribute indexing, typically are not real-time capable, or only support a limited set of attributes. Our approach combines Modifiable Nested Octrees and extended Binned Min-Max Octrees. Using a subset of the well known AHN4 dataset with 138 million points, we evaluate the approach, assess quality and query performance, and compare it with an existing state-of-the-art solution. On commodity hardware, our data structure can process 1.97 million points per second, which is more than most commercially available laser scanners can record. When filtering points by attribute value ranges, it also reduces the number of octree nodes that have to be loaded, and it substantially outperforms naive sequential point filtering.